Yap and its subcellular localization have distinct compartment-specific roles in the developing lung

The generation of organs with complex three-dimensional (3D) tubular architecture and distinct functional compartments, such as the mammary gland, kidney and lung, requires coordinated activation and integration of multiple signaling networks. Yes-associated protein (Yap), a key effector of the Hippo pathway, has been increasingly recognized as a crucial player in this network for its ability to integrate mechanical and chemical signals to control cell behavior during organogenesis (Mauviel et al., 2012; Varelas, 2014; Yu and Guan, 2013). Yap activity is largely dependent on its phosphorylation status, an event controlled by Hippo kinases, including mammalian STE20-like protein kinase (Mst; also known as Stk3) and large tumor suppressor (Lats). Yap phosphorylation on serines (S127 in humans and S112 in mouse) alters its subcellular localization, favoring cytoplasmic retention and, thus, restricting Yap transcriptional output. The nucleocytoplasmic shuttling of Yap is a highly dynamic process, particularly in Yap-expressing structures undergoing rapid morphogenetic changes. However, differential activity of Hippo kinases leads to preferential subcellular localization of Yap in distinct groups of cells or compartments (Manning et al., 2018). This acts as a crucial mechanism to ensure that a proper balance of the Yap transcriptional and cytoplasmic activities is achieved in a particular compartment, preventing Yap-mediated aberrant cell behavior, as observed in cancer (Zhang et al., 2015; Zhao et al., 2011).

Yap and Hippo kinases are expressed in the embryonic lung from early developmental stages in both epithelium and mesenchyme (Lange et al., 2015; Lin et al., 2015; Mahoney et al., 2014). Previous studies demonstrated that Hippo-Yap signaling has an important role in the adult lung, mediating regeneration-repair responses in several injury models (Liu et al., 2016; Otsubo et al., 2017; Volckaert et al., 2017; Zhao et al., 2014). Notably, there is also evidence that this pathway is crucial for lung organogenesis. This was first recognized by studies of severely hypoplastic lungs with truncated airways and distal cysts reported in sonic hedgehog (Shh)^Cre;Yap^flox/flox (Yap^f/f) mutants in which Yap was inactivated in the foregut endoderm at the onset of lung development (Mahoney et al., 2014). Shh^Cre-mediated deletion of Mst1/2 or Lats1/2, leading to the expected nuclear Yap accumulation from the loss of Hippo kinase function, also resulted in disruption of lung development. However, the lung phenotypes differ markedly between these mutants (Lange et al., 2015; Lin et al., 2015; Nantie et al., 2018). Lats1/2-null lungs were severely hypoplastic and had major defects in branching morphogenesis, reminiscent of that seen in Yap-deficient lungs. By contrast, no branching defects were apparent in Mst1/2 mutants, which rather showed abnormal distal differentiation despite displaying the same accumulation of nuclear Yap seen in Lats1/2 mutants.

These observations raised questions about the extent to which these kinases overlap in function and whether the lung abnormalities of these mutants resulted solely from the inability of Yap to shuttle to the cytoplasm or included unrecognized non-Yap-related effects of Mst/Lats deficiency. More fundamentally, information was missing about the role of Yap and its nucleocytoplasmic shuttling in compartment-specific events in the absence of potential confounding effects of Hippo kinase inactivation. Lastly, evidence of the preferential localization of Yap in the nucleus or cytoplasm of the developing lung has been inconclusive because of conflicting results from different reports (Lange et al., 2015; Lin et al., 2015, 2017; Mahoney et al., 2014; Nantie et al., 2018; Szymaniak et al., 2015). Although informative, these studies revealed major gaps of knowledge and left key questions open. Are there specific roles for Yap as the airway and distal lung compartments form and differentiate? If so, do these compartment-specific functions depend on a particular subcellular distribution of Yap in the nucleus or cytoplasm? How would lung development be affected by the inability of Yap to undergo nucleocytoplasmic shuttling in epithelial progenitors?

Here, we provide functional evidence of distinct compartment-specific requirements for Yap in the developing lung. We show that activation of a Yap nuclear transcriptional program is enough to fulfill all Yap requirements essential to form the distal lung. Remarkably, we found that nucleocytoplasmic shuttling is largely dispensable for both branching morphogenesis and sacculation but is crucial for airway differentiation. These observations provide novel insights into the role and regulation of Yap in the developing lung.

To dissect the compartment-specific role of Yap during lung development, we disrupted Yap expression specifically in either distal or (proximal) airway progenitors, once the sex determining region Y-box 9 (Sox9) and the Sox2 domains were clearly established in these compartments. Sox9^CreERT2;Yap^f/f mice were generated and tamoxifen (TMX; 5 mg) was administered sequentially to pregnant mothers at embryonic day (E)11.5 and E12.5. Embryos were analyzed every other day up to E18.5. Morphological analysis revealed no obvious gross abnormalities or differences in body size in mutant mice compared with their no-Yap^f/f control littermates (data not shown). However, lungs from homozygous mutants showed progressively severe cystic expansion of the distal compartment with increasing developmental stage (Fig. 1A). These changes were already noticeable by E14.5 and, by E18.5, the overall morphology of Sox9^CreERT2;Yap^f/f lungs was highly reminiscent of that resulting from global deletion of Yap in Shh^Cre;Yap^f/f mice (Mahoney et al., 2014) as well as from distal Yap deletion using a constitutive Cre (Sox9^Cre;Yap^f/f) reported previously (Lin et al., 2017).

Immunofluorescence (IF) analysis confirmed the selective epithelial inactivation of Yap in the distal cyst-like structures. Although negative for Yap, the epithelium remained strongly labeled for Sox9 (Fig. 1B), further supporting the distal origin of the cysts. Moreover, Yap deletion in the Sox9+ progenitor cell population had no apparent effect on the expression of key early distal cell markers [Sox9, surfactant-associated protein C (Pro-SPC; also know as Sftpc) and bone morphogenetic protein 4 (Bmp4), Fig. S1A], supporting previous observations in Shh^Cre;Yap^f/f mice (Mahoney et al., 2014). By contrast, Sox2 was invariably absent from the Yap-deficient epithelium and was seen only more proximally in Yap-expressing areas unaffected by Sox9^CreERT2-mediated recombination (Fig. 1B,C). Interestingly, Sox2 was clearly detected in the intrapulmonary airways of Sox9^CreERT2;Yap^f/f mutants. However, signals were restricted to the areas of preserved Yap expression, suggesting that the Sox2+Yap+ airway progenitors had been already specified before TMX treatment (E11.5) and, thus, were unlikely to have been generated from Yap-deficient Sox9+ progenitors (Fig. 1B,C). Indeed, TMX administration later in development (E13.5 and E14.5) still resulted in disruption of distal morphogenesis, although the phenotype was much milder: E18.5 lungs only had small distal cysts and mild hypoplasia (Fig. 1D).

We next examined the role of Yap in epithelial buds once Sox2 expression had been established. A previous study proposed Yap to be uniformly active throughout the developing lung epithelium (Lin et al., 2017). Consequently, faulty morphogenesis (as cysts) because of Yap deletion was predicted to occur anywhere along the proximal-distal axis of the developing lung epithelium. We reasoned that, if true, this should be confirmed using a Sox2^CreERT2;Yap^f/f mouse model. Thus, TMX (5 mg) was administered to these mutants at gestational days 11.5, 13.5 and 15.5, and embryos were examined at E18.5 (Fig. 1E). Gross morphology showed that E18.5 Yap-deficient lungs were indistinguishable from those of their Yap^f/f control littermates (Fig. 1E). Efficient targeting was confirmed by the absence of Yap throughout the Sox2+ airway epithelium (Fig. 1F,G). Histological analysis and IF for Sox2/Sox9 showed no obvious defect in branching morphogenesis or proximal-distal patterning in developing airways (Fig. 1H,I; Fig. S1B,C). The distal compartment was histologically comparable with normal saccules populated by both alveolar type 1 and type 2 (AT1, AT2) cells, in sharp contrast to the lungs from Sox9^CreERT2;Yap^f/f mice (Fig. 1H,I; Fig. S1D,E).

Together, these data pointed to a compartment-specific role for Yap in the Sox9+ distal progenitors crucial to generate airways during early lung morphogenesis. This role is no longer seen once Sox2+ airway progenitors are established. Thus, the aberrant lung morphogenesis phenotype common to both Shh^Cre;Yap^f/f and Sox9^CreERT2;Yap^f/f mutants reflects the effect of Yap deficiency specifically in distal epithelial progenitors rather than in all epithelial cells along the proximal-distal axis of the lung.

We further reasoned that, if Yap activity in emerging distal buds was crucial for lung patterning, then the severe disruption of airway morphogenesis seen in Yap-null mice could be corrected by providing nuclear Yap to epithelial progenitors from the earliest stages of lung development. To address this hypothesis, we used a Rosa26-driven human Yap allele (R26Yap^5SA), wherein all serines, normally targeted for Hippo phosphorylation, were mutated to alanines, preventing Yap cytoplasmic sequestration (Cotton et al., 2017). By crossing these mice into Shh^Cre;Yap^f/f lines, we could then examine both the effect of replacing endogenous Yap with a constitutively expressed nuclear Yap (Shh^Cre;Yap^f/f;R26Yap^5SA) and the effect of expressing nuclear Yap in the presence of one endogenous Yap allele as a gain-of-function model (Shh^Cre;Yap^f/+;R26Yap^5SA). Shh^Cre;Yap^f/+ mice reached adulthood and were fertile, as shown in previous reports (Lin et al., 2017; Mahoney et al., 2014).

Gross analysis of E13.5-18.5 Shh^Cre;Yap^f/f;R26Yap^5SA embryos revealed overall morphologically normal lungs with all lobes distinctly formed with none of the highly hypoplastic features or dilated cyst-like structures of the Yap-deficient animals (Fig. 2A-C; Figs S2 and S3B). Yap^5SA-expressing mutants undergoing branching morphogenesis showed nuclear Yap signals on nearly all epithelial cells in both Sox9+ and Sox2+ domains (Fig. S2B,D). Notably, nuclear signals were strong at the transition between distal buds and nascent airways at E13.5 and E15.5 (Fig. S2B,D, insets), suggesting that, in the presence of local nuclear Yap, Sox2+ airway progenitors were specified in the absence of endogenous Yap (Fig. S2B,D). Crucially, a cytoplasmic Yap signal was undetectable in Shh^Cre;Yap^f/f;R26Yap^5SA lungs (Fig. 2D,E), demonstrating that endogenous Yap had been efficiently deleted and replaced by R26Yap^5SA. By E18.5, Shh^Cre;Yap^f/f;R26Yap^5SA whole embryos were smaller (20.3 mm±1.21; P=0.0158) compared with Yap^f/f;R26Yap^5SA controls (23.3 mm±1.27), and could be readily distinguished by external abnormalities, such as a hunched posture, proximally bent thick limbs and a short tail (Fig. S3A,C). Although mutant lungs were also smaller (5.03 mm±0.41; P=0.0051) compared with controls (6.12 mm±0.62), when normalized by the respective whole-body length, the difference became statistically nonsignificant (Fig. S3B,C). IF of phosphorylated Yap (pYap) and Yap confirmed the expected subcellular distribution of signals in control and mutant lungs, and Hematoxylin and Eosin (H&E) staining showed distinct generations of airways and well-formed distal saccules in E18.5 Shh^Cre;Yap^f/f;R26Yap^5SA lungs, compatible with efficient rescue of lung morphogenesis (Fig. 2G,H).

To further investigate a dose-dependent effect, we analyzed the lungs from Shh^Cre;Yap^f/+;R26Yap^5SA mice, which carry an endogenous Yap allele in addition to the Rosa26-driven Yap^5SA construct. Overall morphological features of these lungs were similar to those of Shh^Cre;Yap^f/f;R26Yap^5SA lungs without evidence of disruption of branching morphogenesis or sacculation (Fig. 2C,I), despite the persistence of one endogenous Yap allele, as visualized by faint cytoplasmic Yap stain (Fig. 2F). Thus, although loss of Yap has a profound impact on lung morphogenesis, constitutive expression has no obvious consequences for morphogenesis.

Branching morphogenesis has been traditionally associated with the pseudoglandular stage of lung development (E11-E16), although there is recent evidence that this process extends to the late canalicular stage (∼E17; Alanis et al., 2014). Our previous analysis of Yap expression in the lung focused on earlier stages (E12.5-E14.5; Mahoney et al., 2014). Thus, we expanded this study to later developmental stages using a similar IF protocol and Yap antibodies, which ensured high-resolution analysis of subcellular localization without the artifacts often seen when using signal amplification procedures.

Analysis of E14.5 lungs confirmed the preferentially nuclear Yap localization in nascent Sox9+ distal buds (Fig. 3A). However, at E15.5, cytoplasmic signals became increasingly more noticeable, leading to an overall pancellular Yap staining at the bud tips and neighboring epithelium (Fig. 3B). Interestingly, by the end of the canalicular stage and onset of sacculation (E17-E17.5), the population of Sox9+ distal buds maintained strong Yap cytoplasmic signals but nearly ceased to express nuclear Yap (Fig. 3C). The presence of a strong LPCAT1 signal suggested that these cells were destined to become alveolar type 2 (AT2) cells (Fig. 3D). The immediately proximal Sox2-negative region still expressing pancellular Yap colocalized broadly with Hopx, suggesting that these cells initiate a program of alveolar type 1 (AT1) cell differentiation (Fig. 3E).

Recent studies implicated the appearance of protruding cells in distal buds undergoing sacculation as a mechanism that prevents protruded cells from flattening, thereby precipitating their differentiation into AT2 cells. We investigated the dynamic changes in Yap subcellular localization during these events in Shh^Cre;Rosa26^mTmG mice. IF of GFP, Yap and Pro-SPC in lung sections of these mutants allowed prompt visualization of cell membranes and changes in cell shape in distal progenitors. TMX was administered (50 µg/g) at E15.5 and lungs were examined at E17-E17.5.

Sequential analysis of protrusions in distal epithelial tubules in E17-E17.5 lungs showed the nonoverlapping pattern of Pro-SPC/non-Yap-expressing cells (prospective AT2 cells) and the Pro-SPC-negative/Yap-expressing cells (Fig. S4A). As the lumen of the regions proximal to these protrusions enlarged, nuclear Yap+ cells appeared to flatten, whereas Yap-negative Pro-SPC-expressing cells that maintained minimal contact with the luminal surface retained their cuboidal shape (Fig. S4B,C). These patterns persisted in the E18.5 distal lungs (Fig. 4A; Figs S4 and S5), where IF-confocal imaging revealed cells expressing nuclear or pancellular Yap double-labeled with AT1 markers (T1a, Ager and Hopx) (Fig. 4A; Fig. S5, control). By contrast, Yap-negative cells consistently expressed markers of AT2 cells (Abca3, Pro-SPC, SPB and LPCAT1) (Fig. 4A; Fig. S5, control).

These intriguing patterns led us to examine, first, how local Yap affected specifically these distal events during late development regardless of its subcellular localization. Our analysis of Sox9^CreERT2;Yap^f/f mice showed that, despite its aberrant, cyst-like appearance, the early identity of the distal epithelium (Sox9, Pro-SPC and Bmp4) was unaffected (Fig. S1A). However, it was unclear whether the progenitors in these cysts were still able to properly respond to signaling cues for late differentiation. Remarkably, despite the highly distorted morphology of distal airspaces, E18.5 Sox9^CreERT2;Yap^f/f lungs were extensively populated by AT2 cells, as demonstrated by the large number of cuboidal cells expressing multiple AT2 markers (Pro-SPC, SPB, LPCAT1 and Abca3; Fig. 4A; Fig. S5). By contrast, at late gestation, Yap-deficient Sox9 progenitors in these cysts were unable to turn on the AT1 markers T1a, Ager or Hopx (Fig. 4A; Fig. S5). Quantitative analysis confirmed the significant decrease in the percentage of AT1 and increase in the AT2 cell population (P=0.00169; Fig. 4B). We concluded that, although dispensable for AT2 cells, Yap is crucial for initiation of the program of AT1 cell differentiation.

Next, we examined how preventing the dynamic changes in Yap subcellular localization observed before sacculation would interfere with late distal differentiation. What if Yap was available but unable to undergo nucleocytoplasmic shuttling? The overall rescue of branching morphogenesis and sacculation of our Shh^Cre;Yap^f/f;R26Yap^5SA mutants in the absence of Yap nucleocytoplasmic shuttling provided an opportunity to address this question. Analysis of E18.5 mutant lungs showed nuclear Yap in both AT1 and AT2 cells, in contrast to control (Yap^f/f;R26Yap^5SA) littermates (nuclear Yap was only seen in AT1 cells) (Fig. 4C). We asked whether the presence of nuclear Yap (R26Yap^5SA) in distal epithelial buds at the onset of AT1 versus AT2 differentiation and the inability of Yap to be sequestered to the cytoplasm influenced distal cell fate decisions. Morphometric analysis showed a slight decrease in the proportion of AT2 cells (42.1±2.3%; P=0.0321), but no significant change in that of AT1 cells (48.7±3.2%; P=0.965) in mutants (Fig. 4D). Interestingly, maintained nuclear Yap expression led to a significant increase in the fraction of double Pro-SPC+T1a+ cells (9.2±1%; P=0.00134; Fig. 4D,E) relative to controls (2.7±0.4%).

Thus, expression of Yap is crucial in distal progenitors to allow morphogenesis and differentiation of the distal lung; this function appears to be minimally affected by the inability of Yap to undergo nucleocytoplasmic shuttling.

Yap has been shown to have a significant impact on airway epithelial differentiation of the adult lung, using both in vivo and in vitro methodologies (Lange et al., 2015; Lin et al., 2015; Mahoney et al., 2014; Otsubo et al., 2017; Szymaniak et al., 2015; Zhao et al., 2014). However, its role in the differentiation of developing airways remained elusive and could not be adequately concluded from the mouse genetic models reported. Difficulties included the severe arrest in airway morphogenesis with accumulated defects caused by early Yap inactivation in the embryonic lung, and efficiency of recombination in some of the mouse Cre lines reported (Lin et al., 2017).

The lack of morphogenetic abnormalities and preserved 3D structure of the airways of Sox2^CreERT2;Yap^f/f mice provided an opportunity to explore this issue. E18.5 lungs from these mutants were examined by IF/confocal microscopy for the distribution and levels of epithelial markers of multiciliated and secretory cell differentiation. Both cell types were abundantly present in the airways of these mutants, although expression of CC10, which marks the secretory Club cells, was weaker. Quantitative analysis revealed a decrease in the proportion of CC10-expressing cells (P=0.0103), but an increase in that of the multiciliated cells [Foxj1: P=0.0154; acetylated alpha-tubulin (ac-aTub): P=0.00626] (Fig. 5A-D). These observations and the presence of epithelial Yap largely in the cytoplasm of control airways suggested that, during development, Yap is unlikely to specify airway epithelial cell lineages. Instead, it could act in a mechanism influencing the differentiation of these cell types. Although not explored here, previous studies suggest a role for Yap in ciliogenesis and in the regulation of Notch, a known determinant of cell fate decisions in the airway epithelium (de Lima et al., 2016; Grampa et al., 2016; Kim et al., 2015; Mahoney et al., 2014; Totaro et al., 2017; Tschaharganeh et al., 2013; Tu et al., 2018; Yimlamai et al., 2014).

To gain additional insights into the role of Yap in developing airways, we investigated the lungs in which epithelial Yap was maintained constitutively active from the onset of lung development. Although Yap was clearly found in the cytoplasm of branching epithelial tubules, it was unclear whether cytoplasmic sequestration of Yap was required to carry its functions in developing airways. Thus, we examined the airways of Shh^Cre;Yap^f/f;R26Yap^5SA mice in which cytoplasmic Yap sequestration was abolished and Yap subsequently accumulated in the nucleus. Analysis of E18.5 mutants showed a highly disorganized epithelium in trachea, extrapulmonary bronchi and intrapulmonary airways (Fig. S6A,B). Although preserving its pseudostratified organization, multiple clumps of cells strongly labeled by nuclear Yap appeared in the lumen, detached from the airway epithelium. This phenotype was reminiscent of that reported in Lats1/2-null, Crb3-null and other mouse lines that result in Yap constitutive nuclear expression (Lange et al., 2015; Lin et al., 2015; Mahoney, 2014; Nantie et al., 2018; Otsubo et al., 2017; Szymaniak et al., 2015; Zhao et al., 2014). Interestingly, this hyperplastic phenotype was not associated with a significant increase in cell proliferation (Fig. S6D), because quantification of 5-ethynyl-2´-deoxyuridine (EdU)+ epithelial cells showed that EdU incorporation in R26Yap^5SA+ lungs was not significantly different from controls (P=0.317; Fig. S6D).

We asked whether the detached luminal cell clumps represented apoptotic cells, given that nuclear Yap has a role in anoikis (Zhao et al., 2012). IF analysis of Cleaved Caspase 3 (CC3) showed no evidence of apoptosis in the epithelium or in the detached clumps of cells (Fig. S6E). Intriguingly, the luminal clumps of detached cells that expressed strong nuclear Yap were also strongly labeled with T1a, suggestive of the AT1 fate (Fig. S6E). We asked whether these abnormalities could be exacerbated by an additional Yap allele in our gain-of-function model. Interestingly, airways from E18.5 Shh^Cre;Yap^f/+;R26Yap^5SA lungs were very different from those of Yap^f/f;R26Yap^5SA controls but nearly indistinguishable from Shh^Cre;Yap^f/f;R26Yap^5SA lungs (Fig. S6C). No significant difference in Edu incorporation was detected between control and Shh^Cre;Yap^f/+;R26Yap^5SA lungs (P=0.608; Fig. S6D). Thus, we were unable to ascribe the appearance of the luminal clumps in mutant airways to excessive cell proliferation or cell death.

We then examined how the replacement of endogenous Yap with a constitutively active nuclear Yap, unable to shuttle to the cytoplasm from the earliest lung developmental stages, influenced airway epithelial differentiation. Confocal analysis of E18.5 Shh^Cre;Yap^f/f;R26Yap^5SA lungs identified nuclear Yap at distinct levels (hereafter, Yap^high, Yap^mid or Yap^low) in the airway epithelium colocalizing with differentiation markers. The patterns of colocalization were determined for each marker and quantitated by Fluorescence Intensity Profile Analysis (see Materials and Methods; Fig. S7A-D). Yap^high cells did not colocalize with any of the markers tested, suggesting incompatibility with airway differentiation (Fig. 6A-D; Fig. S7E). In Yap^mid cells, only ∼10-15% expressed Scgb3a2 or Foxj1, known as early markers of secretory and multiciliated cells, respectively (Fig. 6A,D; Fig. S7E). Surprisingly, CC10+ secretory cells colocalized with intermediate levels of nuclear Yap (35.6±4.2%; Fig. 6B,D). Although Scgb3a2 and CC10 are expressed in secretory cells, these markers did not appear to label the same cell population (Fig. S7F). p63+ basal cell precursors did not colocalize with Yap^high, although some did with Yap^mid (34±8.4%; Fig. 6C,D). Together, these data suggest that high levels of nuclear Yap abrogate airway lineage differentiation.

Intriguingly, none of the major airway lineage markers was present in the cells efficiently expressing high levels of nuclear Yap (R26Yap^5SA). We asked whether, despite acquiring an airway fate earlier (Fig. S2), this fate was maintained with continued constitutive nuclear Yap expression. Thus, we stained E18.5 Shh^Cre;Yap^f/f;R26Yap^5SA lungs with Sox2 and Yap and quantified the proportion of double-labeled cells in the different populations according to the intensity of the fluorescent Yap signals, as before. Surprisingly, only a few Sox2+ cells were seen in the Yap^high population (2.3±1.63%), whereas most Sox2+ cells expressed only weak nuclear Yap (Yap^low=74.1±5.9%) (Fig. 7A). This suggested that, overall, Yap^high cells could have lost their airway identity, reverting to an earlier distal progenitor state. This could explain the inability of these cells to turn on airway-specific programs of differentiation. Therefore, we stained these lungs with distal markers of early (Sox9 and Pro-SPC) and late (T1a, Hopx and Ager) cell fate. Although no Sox9 or Pro-SPC labeling was present, T1a, Hopx and Ager were strongly expressed in Yap^high airway epithelial cells in which Sox2 was inhibited (Fig. 7B,D; Fig. S7G), and T1a colocalized with Yap^mid airway cells (Fig. S7H). The nonoverlapping pattern of T1a with p63 ensured that the T1a-expressing Yap^high cells were not airway basal cells but indeed type 1 cells (Fig. 7C). Together, these data indicated that, although activation of transcriptional programs by nuclear Yap is essential for early morphogenesis, Yap nucleocytoplasmic shuttling is crucial in controlling these programs by bringing Yap preferentially to the cytoplasm in forming airways. This helps to consolidate proximal-distal fate in these compartments, allowing initiation of cell type-specific differentiation and ensuring that distal programs (such as the AT1 differentiation) do not occur in airways. These observations highlight the significance of the dynamic subcellular expression of Yap when airways are forming and during the transition from the canalicular to saccular stage reported here.

Here, we provide novel evidence of compartment-specific roles for Yap in lung morphogenesis and epithelial differentiation and how Yap subcellular localization influences these events. Lineage-tracing studies demonstrate that epithelial progenitors at the bud tips (Sox9+ and Id2+) are multipotent and leave the distal domain of branching tubules to give rise to all epithelial cell types of the airway and alveolar compartments (Rawlins et al., 2009; Yang et al., 2018). We found that, when Yap expression was disrupted selectively in Sox9+ progenitors of the developing lung, cells of the distal compartment were no longer able to respond to local cues to initiate a program of airway morphogenesis and ultimately form distal and proximal branches. Rather, Sox9^CreERT2;Yap^f/f mice formed distal cystic structures, a phenotype essentially similar to that reported in Shh^Cre;Yap^f/f mice and supported by studies in Sox9^Cre;Yap^f/f mice (Lin et al., 2017), suggesting that the lung defect in Shh^Cre;Yap^f/f mice results primarily from disruption of Yap-mediated events in Sox9+ progenitors. Remarkably, restoring Yap distal epithelial activity in a cell autonomous fashion and solely in the nucleus enabled cells to respond to these cues, allowing epithelial-mesenchymal cross talk for proper morphogenesis. Although our data suggest that Yap activity in distal progenitors is important for morphogenesis, we cannot rule out cell nonautonomous effects through altered epithelial-mesenchymal crosstalk, in the absence of a lineage tag. This is similarly true for our experiments wherein Yap was deleted in airway progenitors.

Under physiological conditions, nucleocytoplasmic shuttling of Yap is expected to occur in a dynamic fashion throughout all Yap-expressing cells of a developing structure. Nevertheless, several reports also show that, despite being continuously shuttled, the subcellular localization of Yap/Yorkie differs even within cell populations of the same organ, influenced by variables such as the stage in the cell cycle and others (Manning et al., 2018; Varelas, 2014). Indeed, previous reports demonstrated differences in Yap subcellular localization in the developing lung, preferentially nuclear in distal buds and cytoplasmic in forming airways (Mahoney et al., 2014; Szymaniak et al., 2015). However, less stringent staining protocols showing a rather pancellular distribution of Yap argued that any given region of the lung epithelium could be susceptible to loss of Yap and, thus, produce local branching defects (Lin et al., 2017). Here, we demonstrated that this is not the case, because no branching abnormalities were observed by deleting Yap in airway progenitors once Sox2 expression was established (Fig. 1E). Moreover, we provide new evidence of striking differences in Yap subcellular localization arising in the distal lung at the onset of sacculation. These dynamic patterns culminate with nuclear Yap-expressing cells differentiating into AT1 cells, in stark contrast to AT2 cells, in which Yap is largely cytoplasmic. Lastly, the distinct effects of nuclear Yap in Shh^Cre;Yap^f/f;R26Yap^5SA mice rescuing Yap-dependent events in distal progenitors (branching and sacculation) but disturbing the integrity and differentiation of the airway epithelium, do not support the idea that proximal and distal compartments respond to Yap in the same fashion. This is consistent with our observation of distinct subcellular Yap localization in these compartments, rather than a model in which Yap is equally distributed in the nucleus and cytoplasm throughout the developing lung epithelium.

Remarkably, nuclear-cytoplasmic shuttling appears to be dispensable for all events associated with the distal lung, provided that nuclear Yap is available to mediate transcriptional events in distal progenitors. The steady-state levels of Yap expression driven by the Rosa locus in Shh^Cre;Yap^f/f;R26Yap^5SA mice from the earliest developmental stages were sufficient to rescue the Yap-dependent signaling of all key events of the developing lung, including branching morphogenesis, sacculation and epithelial differentiation. Thus, expression of a Hippo-independent Yap can substitute for endogenous Yap and exert its function in the distal lung.

It was surprising that in our gain-of-function Shh^Cre;Yap^f/+;R26Yap^5SA model, the activity of an additional Yap allele had no notable increased effect in any of the lung compartments compared with the Shh^Cre;Yap^f/f;R26Yap^5SA model. We reasoned that the abnormal ectopic AT1 and disruption of the airway epithelial integrity in these mutants was probably a function of the continued expression of Yap in the nucleus rather than an increase in Yap dosage. This is also consistent with observations from other genetic models in which Hippo components and effectors, such as Mst1/2 (Lange et al., 2015; Lin et al., 2015), Last1/2 (Nantie et al., 2018) and Crb3 (Szymaniak et al., 2015), were inactivated.

Notably, in the absence of nucleocytoplasmic shuttling, constitutively active nuclear Yap inhibited airway progenitor cell identity, as demonstrated by the low Sox2 expression in high Yap-expressing cells. Indeed, Yap regulation of Sox2 expression has been previously reported and further confirmed in the present work by chromatin immunoprecipitation (ChIP) analysis in Sox2-expressing airway progenitors (Mahoney et al., 2014; Fig. S7I). Interestingly, neither bud morphogenesis nor initiation of the patterning events that give rise to airways was disrupted in these mutants despite the maintained expression of nuclear Yap and lack of cytoplasmic Yap. Moreover, our analysis of Shh^Cre;Yap^f/f;R26Yap^5SA and Shh^Cre;Yap^f/+;R26Yap^5SA mice showed that the differentiation program of the distal lung epithelium was remarkably resistant to the maintained expression of constitutively active Yap. This did not prevent AT2 cells from forming and had only a relatively mild effect on the balance of AT1×AT2 distal cells, unlike a previous study, which showed an increase in the number of AT1 at the cost of AT2 cells (Nantie et al., 2018). The difference in findings could be explained by the distinct genetic model used. Whereas we had Yap^5SA under the control of the Rosa26 locus, Nantie et al. used a doxycycline-inducible Sftpc promoter construct, which drives high levels of Yap^S127A expression throughout lung development. This could have allowed a more rapid and efficient accumulation of Yap^S127A, leading to a severe disruption of the balance of AT1×AT2 cells. Nevertheless, both studies observed ectopic AT1 cells in airways as a result of forced nuclear Yap, suggesting that Yap is involved in the differentiation of AT1 cells.

Together, these findings emphasize the requirement of a mechanism of nucleocytoplasmic shuttling to foster Yap cytoplasmic sequestration in nascent airway epithelial progenitors. This would consolidate the proximal fate, allow compartment-specific initiation of cellular differentiation and maintain integrity of the architecture of the pseudostratified epithelium.

All studies were approved by the Columbia University Irving Medical Center Institutional Animal Care and Use Committee (IACUC). Shh^Cre (Harfe et al., 2004), Sox9^CreERT2 (Soeda et al., 2010), Sox2^CreERT2 (Arnold et al., 2011), Yap^f/f (Camargo et al., 2007), R26Yap^5SA (Cotton et al., 2017) and Rosa26^mTmG (Muzumdar et al., 2007) lines have been described previously. To disrupt Yap specifically in the developing distal or airway epithelial compartments, Sox9^CreERT2 and Sox2^CreERT2 lines were crossed with Yap^f/f lines. To generate mice in which endogenous Yap was replaced by constitutively active nuclear Yap, Shh^Cre;Yap^f/+ lines were crossed with R26Yap^5SA mutants. TMX (Sigma, T5648) was dissolved in sunflower seed oil (Sigma, S5007) at a stock concentration of 20 mg/ml. The various TMX-administration regimens are described in the results section. Detection of a vaginal plug was considered to be E0.5. All mouse lines were generated in a B6/C57 background, except the R26Yap^5SA line (originally CD1). Sox9^CreERT2 and Sox2^CreERT2 lines were a gift from Dr Jianwen Que (Columbia University Irving Medical Center), the Yap^f/f line from Dr Fernando Camargo (Harvard University) and the R26Yap^5SA line was from Dr Junhao Mao (University of Massachusetts). For crosses that involved the Shh^Cre allele, matings were set up such that only the male was to carry this allele. For other matings, there was no preference in bearer of the Cre construct.

Embryonic lungs at the developmental stages indicated were dissected, rinsed in PBS and fixed in 4% paraformaldehyde in PBS overnight at 4°C. Samples were then processed for frozen (optimal cutting temperature; OCT) embedding. H&E staining was performed by the Columbia University Irving Medical Center, Herbert Irving Comprehensive Cancer Center Molecular Pathology (histology) core, and imaged by Leica SCN400 Slide Scanner. Immunofluorescent staining was performed on 5-8-µm-thick tissue sections. Slides were subjected to heat-based antigen retrieval using a citrate-based buffer (Vector Labs, H-3300) for 8:30 min and blocked at room temperature (RT) for 1 h in 5% donkey serum in PBS. Staining with mouse primary antibodies was performed using MOM kit (Vector Labs, BMK-2202). Sections were incubated overnight with primary antibodies at 4°C. The following day, slides were washed, incubated for 2 h with Alexa Fluor-conjugated secondary antibodies as well as NucBlue Fixed Cell Stain ReadyProbes (DAPI) (Invitrogen, R37606) diluted in 5% donkey serum, washed again and finally mounted using Prolong Gold anti-fade (Invitrogen, P36930) reagent and imaged on a Leica DMi-8 Deconvolution microscope, and a Zeiss LSM 710 confocal microscope system.

The following primary antibodies were used: rabbit anti-Abca3 (1:1000, Seven Hills, WRAB-70565), rabbit anti-acetylated α-tubulin (1:3000, Cell Signaling, 5335S), mouse anti-acetylated α-tubulin (1:3000, Sigma, T7451), rabbit anti-Ager (1:100, AbCam, ab3611), goat anti-CC10 (1:100, Santa Cruz, sc-9772), rabbit anti-cleaved caspase 3 (1:100, Cell Signaling, 9661P), mouse anti-FoxJ1 (1:100, eBioscience, 14-9965-82), rabbit anti-Hopx (1:100, Santa Cruz, SC-30216), rabbit anti-Nkx2.1 (1:1000, Seven Hills, WRAB-TTF1), rabbit anti-p63a (1:100, Cell Signaling, 13109S), rat anti-scgb3a2 (1:100, R&D Systems, MAB3465), rat anti-Sox2 (1:200, eBioscience, 14-9811-82), goat anti-Sox9 (1:200, R&D systems, AF3075), rabbit anti-pYap (1:50, Cell Signaling, 4911S), rabbit anti-Pro-SPC (1:1000, Seven Hills, WRAB-76694), hamster anti-T1a (1:50, Developmental Studies Hybridoma Bank #8.1.1, RRID:AB_531893), rabbit anti-Yap (1:100, Cell Signaling, 14074S) and mouse anti-Yap (1:100, Santa Cruz, sc-101199).

The following secondary antibodies were used: donkey anti-goat (conjugated with Alexa Fluor 568 or 647), goat anti-hamster (conjugated with Alexa Fluor 647), donkey anti-mouse (conjugated with Alexa Fluor 488, 568 or 647), donkey anti-rabbit (conjugated with Alexa Fluor 488, 568 or 647), donkey anti-rat (conjugated with Alexa Fluor 488 or 647) and donkey anti-chicken (conjugated with Alexa Fluor 488). All secondary antibodies were obtained from either Jackson ImmunoResearch or Fisher Scientific.

In situ hybridization was performed as described before (Chen et al., 2007). Briefly, probes were synthesized by PCR from mouse genomic cDNA. The primers (including T3 and T7 sequences) were as follows: forward primer: AATTAACCCTCACTAAAGGGAAGCTAGGTGAGTTCGGCATC; reverse primer: TAATACGACTCACTATAGGGGCTCCTAGCAGGACTTGGCAT. Frozen sections were subjected to heat-based antigen retrieval in a citrate-based buffer (Vector Labs, H-3300), acetylated, permeabilized in PBST and prehybridized in prehybridization buffer for 2 h. Hybridization with a Bmp4 probe proceeded overnight at 72°C in a humidified chamber. The following day, slides were rinsed in SSC buffer, blocked in 2% sheep serum and incubated in anti-digoxigenin-AP Fab fragments (1:2500, Roche, 11093274910) overnight at 4°C. The next day, the slides were rinsed in PBST and developed in BM Purple AP Substrate (Roche, 11442074001). Slides were then imaged on a Leica DMi-8 Deconvolution microscope.

Cell proliferation was assessed in control and R26Yap^5SA-expressing Shh^Cre;Yap^f/f or Shh^Cre;Yap^f/+ mice by exposing embryos in utero to 1 mg EdU via intraperitoneal injection in mothers at gestational day E17.5 (5 mg/ml stock solution in PBS). Embryos were harvested 24 h later and processed for OCT embedding and sectioning. EdU was detected by immunostaining using the Click-iT EdU Alexa Fluor 647 Imaging kit (Thermo Fisher Scientific, C10340) following antigen retrieval. Sections were subsequently co-stained for Yap as described above.

Yap recruitment and activation of transcription at the Sox2 locus was examined by ChIP-PCR in airway epithelial progenitors (basal cells) isolated from adult mouse trachea as described previously (Mahoney et al., 2014). Briefly, basal cells were expanded in submerged cell culture conditions for 5 days (You et al., 2002), then washed, fixed and sonicated, such that ∼200-bp DNA fragments were obtained. Samples were incubated with rabbit anti-Yap (Cell Signaling, 14074S) or IgG (Cell Signaling, 3600S) primary antibodies (2.5 µg/ml) overnight at 4°C. The next day, samples were incubated with Dynabeads Protein G Immunoprecipitation Kit (Invitrogen, 100.07D) for 2 h at 4°C and washed with NaCl overnight at 65°C. Following uncrossing of the DNA, DNA was purified using a Qiagen Qiaquick PCR purification kit (Qiagen, 28104). PCR of Sox2 was carried out on genomic DNA, ChIP input DNA and DNA pulled down by either Yap or IgG antibodies, using three primer sets: negative control (non-Yap-TEAD-binding site upstream of Sox2 locus), primer pair 1 (predicted Yap-TEAD-binding site in Sox2 promoter), and primer pair 2 (TEAD-binding sequence within Sox2-coding region) (Fig. S7I). Given that primer pair 2 spanned a region within the Sox2-coding region, no TEAD binding was expected and, therefore, only primer pair 1 was hypothesized to generate a band. The following primers were used: negative control forward primer: TCTGCTGTTGGCTTTCATTG; reverse primer: GGTAACCGTTGTTGGGTTGT. Primer pair 1 forward primer: ATTGAGTTATCAAGGCAGTA; reverse primer: GCAGTAATTAGCGAGAAGTA. Primer pair 2 forward primer: GTTGTCCACATCAACTCGTT; reverse primer: CGCCAGATAAGTGGGAGGTT.

Shh^Cre;R26^mTmG embryos were dissected at E16.5 and transferred to a 60-mm dish coated with silica gel. The four limbs of each embryo were affixed to the silica gel using needles. Ribs and sternum were removed to expose the heart and lung. A 30-gauge needle was inserted into the right ventricle. DMEM/F12 medium supplemented with 30% FBS and 1 U/ml penicillin-streptomycin (Thermo Fisher Scientific) was pumped (BT100-2J, Longer) into the heart at the speed of 50-60 µl/min. To mimic fetal breathing movements, a 30-gauge needle was inserted into the trachea and PBS was pumped into and out of the trachea at a speed of 6-8 µl/10 s (each time). The frequency of these fetal breathing movements was 6 breaths/min.

All image-based quantification analyses were performed in ImageJ (version 1.50i) (Schneider et al., 2012), using plugin BioFormats (Linkert et al., 2010). Statistical analysis was performed in Microsoft Excel 2016 or R (version 3.4.3, R Core Team, 2014), running in RStudio (version 1.1.383, RStudio, 2013), using R package ggplot2 (Wickam, 2016). For comparison of two groups, statistical significance was determined by a two-sample two-tailed t-test (Figs 4B and 5D). For comparison of three groups or more, a one-way ANOVA followed by a post-hoc Tukey test was used (Figs 4D, 6D, 7A; Figs S3C, S6D, S7E). N values can be found in figure legends and represent the number of animal samples used per genotype. Data in graphs are presented as a boxplot with median, minimum, 1st and 3rd quantiles, and the maximum value, as indicated in figure legends. Statistical significance was defined as P<0.05.

For the measurement of lung length of R26Yap^5SA+ litters, photographs were taken of the right lateral side of each embryo, as well as ventral photographs of the respective lungs. Body length (mm) was obtained from measurements using a ruler of the head-to-tail distance of the embryo. Lung length (mm) was obtained similarly by measuring the distance between the upper and lower edge (anterior-posterior axis) of the left lungs. Lung-to-body length ratios were calculated by dividing lung length by body length. For both full-body as well as lung lengths, control n=13, Shh^Cre;Yap^f/f;R26Yap^5SA n=4 and Shh^Cre;Yap^f/+;R26Yap^5SA n=8.

Quantification of the number of CC10+, FoxJ1+ or ac-aTub+ cells in E18.5 Sox2^CreERT2;Yap^f/f lungs was performed in three control and three mutant samples. Three nonconsecutive sections were stained per sample for each marker, after which the number of marker+ cells was counted. Cell counts were normalized per sample by dividing the number of marker+ cells by the total cell count to obtain a per sample percentage of marker+ cells contributing to the airway epithelium. Data were presented in boxplots (Fig. 5D).

To determine the proportion of epithelial cells expressing Yap in various cell types, nonconsecutive sections from E18.5 Shh^Cre;Yap^f/f;R26Yap^5SA lungs (n=3) were co-stained with Yap and each of the following cell differentiation markers: T1a (alveolar type 1), Scgb3a2 (secretory), CC10 (secretory Club), p63 (pre-basal), FoxJ1 (early multiciliated) or Sox2 (airway progenitor). Nine sections were stained per marker deriving from three biological samples and photographed on a Leica DMi-8 deconvolution microscope. Despite being expressed from a Rosa26 promoter, the intensity of Yap5SA signals varied among airway progenitors. Profile analysis of signal intensity across a line (in µm) was performed in representative areas of Yap double-labeled sections, using ImageJ macro RGB_Profile_Plot (Fig. S7A-D). The signal intensity range (0-255) of Yap was then classified into three categories (0-85, 85-170 and 170-225), which were used to define low, intermediate (mid) and high Yap signal levels, respectively (Fig. S7A-D). Colocalization of the various markers with each of these Yap signal level categories was then quantified and represented as a percentage in graph form (Fig. 6D; Fig. S7E,H).

Quantification of proliferating cells was performed in three nonconsecutive sections of E18.5 control (n=3), Shh^Cre;Yap^f/f;R26Yap^5SA (n=3) and Shh^Cre;Yap^f/f;R26Yap^5SA (n=2) EdU-labeled sections. Confocal microscopy images were imported into ImageJ, Nkx2.1/Sox2 and EdU channels were colored red and green, respectively, and double-labeled cells (appearing yellow) were highlighted and counted by consecutively using the (color) threshold and particle analysis functions. The automated count of the cells was manually checked for accuracy. Values for each group were summed per sample to obtain average numbers of proliferating cells and presented as percentage of EdU+ epithelial cells and represented in boxplot graphs (Fig. S6D).

Quantification of the number of AT1 and AT2 cells in control, Sox9^CreERT2;Yap^f/f and Shh^Cre;Yap^f/f;R26Yap^5SA lungs (n=3 per genotype) was performed in three nonconsecutive sections by staining for T1a, Pro-SPC and Yap followed by confocal microscopy to image up to ten nonoverlapping fields per section. The number of T1a+ and Pro-SPC+ cells was manually counted using the ImageJ Cell Counter until a total of ∼1000 cells was counted per section. The percentage of AT1 and AT2 cells was derived by dividing the number of counted cells by the total number of cells counted per section, after which the percentages were averaged per sample and represented in boxplot graphs (Fig. 4B,D).

We would like to thank Xaralabos Varelas, Munemasa Mori, Maria Stupnikov and Ying Yang for technical advice and thoughtful discussions, and Dr Jianwen Que and Dr Junhao Mao for generously providing the Sox9^CreERT2 and Sox2^CreERT2, and R26Yap^5SA mouse lines, respectively.